Non-Volatile Memory Express (NVMe) is the Future of Flash Storage
Businesses have to manage vast and incessantly-growing volumes of information driven by adoption of data-intensive technologies, which continue to generate ever-more parallel data streams. The new NVMe host controller interface and storage protocol can process up to 64,000 parallel data streams – compare that with the SATA or SAS protocols, which can only handle one queue at a time. Fujitsu has already announced plans to adopt native NVMe combined with PCIe connected solid-state drives in its future storage product lines alongside our ETERNUS AF ALL-Flash Arrays, providing massively parallel, previously unattainable data transfer speeds. In 2019, this enormous boost in bus speeds will enable customers to manage massively parallel data access while guaranteeing minimum latency. While NVMe ultra-high-speed interfaces primarily exist “within the box”, with NVME over Fabric (NVMe-oF) the potential impact of NVMe is no longer limited to just accelerated SSDs – it will power new performance thresholds in servers, fabrics, storage, and system management.
We’ll see Quantum computing and the Fujitsu Digital Annealer take off
From an arcane idea just a few years ago, quantum computing is making rapid progress and represents the future of computing, without a doubt. But something for you to consider next year? Really? Well, yes. Fujitsu’s novel approach – the Digital Annealer, a quantum-inspired technology-based solution – delivers the ability to solve combinatorial optimization problems that would simply not be possible using conventional computer architectures, at least within the lifetime of a human being. This includes solving challenges such as the search for similarities in molecules for drug or new material discovery, portfolio risk optimization in financial services -safeguarding from a potential financial crash, increasing robot production efficiency, warehouse optimization and reducing production time as part of factory optimization, and real-time traffic route optimization, enabling public transportation planning.
Far from being a theoretical experiment, Fujitsu Digital Annealer is already working to help customers rapidly solve such challenges such as searching for molecular similarities in drug research, reducing production times for custom manufacturing and optimizing the arrangement of warehoused components for factories and logistics. In October, we announced details of a ground-breaking proof of concept (PoC) project with NatWest in the UK to solve some of the bank’s most complex, challenging and time-consuming financial investment problems. This is through optimizing its mix of high-quality liquid assets including bonds, cash, and securities. Maybe in 2019, you could be tackling computational challenges that you thought were intractable?
Machine learning will be accelerated by digital annealing
Building on the quantum-inspired approach, it is machine learning where all the hard work is done in creating an Artificial Intelligence. The process starts by teaching an AI. If you want it to identify a cat, you need to show it millions of pictures of cats. This process takes time, and if it doesn’t work properly, you need to start over. The Fujitsu Digital Annealer can help accelerate this training process, for example by offering partially pre-trained components such as speech synthesis recognition.
A new type of memory will revolutionize server workloads
NVMe technology not only influences the future of flash storage but is also playing an essential role in the development of a future generation of servers. Fujitsu is Intel’s cooperation and development partner, responsible for developing the prototype server system based on the Intel Optane DC persistent memory, within the EU Horizon 2020 project called NEXTGenIO. This prototype system, which is based on PRIMERGY servers, leverages ‘non-volatile memory with persistence’, which unlike traditional DRAM combines the speed of main memory with the persistence of conventional storage. By enabling affordable system memory capacities with this new class of memory, end customers can better optimize their workloads by moving and maintaining larger amounts of data closer to the processor, minimizing the higher latency of fetching data from system storage. This new technology will be available in the upcoming generation (M5) of the PRIMERGY dual- and quad-socket as well as our PRIMEQUEST 3800 series 8-socket servers, which are planned to be launched in April 2019.
Containerization will be an enabler for hybrid IT use cases
We have already heard a lot about containers and 2019 will only see that emphasis increase. Containerization is an aspect of the so-called “micro-service” approach, where you decompose an application into small, atomic, encapsulated entities with a limited functional scope. These containers can then be dynamically re-composed into complex applications and distributed to various platforms, allowing concurrent processing of multiple tasks. The payback is that micro-services simplify development (usually enabling super-agile DevOps) and maintenance, improve responsiveness to changing demands, enable scalability at individual service levels and contribute to cost-effectiveness.
And because containers make it easier to focus on new features, fixing issues and shipping new or updated software, they are vehicles for delivering customer value. A handsome return! We deploy containerization where it meets the demands of use cases, via open source software, such as OpenStack, Docker, and Kubernetes, as well as via in VMware and Microsoft environments, supplemented by Fujitsu’s upper layer software products, such as ServerView Cloud Monitoring Manager and Enterprise Service Catalog Manager. And, as a multi-cloud service provider, we deploy it also on our partner clouds, such as AWS and Microsoft Azure.
Liquid Immersion Cooling Systems will make a breakthrough
Getting the most out of AI or Big Data requires ever-greater performance from IT systems. Accelerator chips such as GPUs require ever greater amounts of energy for power, resulting in a dramatic rise in operating temperatures, and the requirement for more cooling. In some cases, data center air conditioning systems are already inadequate, which puts limits on how we can exploit the full potential of high-end systems. Fujitsu’s Liquid Immersion Cooling Technology is a solution: it uses efficient, fluid-based cooling, reducing the demand for large-scale air-conditioning equipment and reducing total power consumption by around 40 percent. This reduction in power consumption means customers in data centers with capped power supplies can approximately double their server density and therefore their compute power. Either way leads to big savings in 2019.